Attending GECCO 2017 in Berlin this July was a real treat, an opportunity to immerse myself for five full days in all things genetic and evolutionary computation, and to catch up with the over 600 delegates. Having not personally attended a GECCO since Amsterdam in 2012, it is inspiring to see so many new researchers enriching our community, and a new sense of energy and urgency with the current wave of interest in all things Artificial Intelligence. Of course we all know, Artificial Intelligence != Deep Learning, mind you Deep Learning got a mention at almost every single session I attended!
GECCO 2017 was in part a sentimental journey for me. It put into focus my personal journey as a researcher in evolutionary computation who over the past five years has been dedicated to, and distracted by, senior management roles in academia, in addition to the pleasure of being a husband and a father to three wonderful young children. The knock on consequence has resulted in less EC conference travel since 2012.
My personal GECCO journey began almost twenty years ago pre-GECCO in 1998 when I attended the First European Workshop on Genetic Programming in Paris, and GP 98 in Madison, Wisconsin. These were my first conferences, with my first ever presentations, and marked the first publications on my favourite approach to Genetic Programming, Grammatical Evolution. What stood out to me from those early events was the collegiality of our community. As a fresh-faced and intimated PhD student I have a very fond memory of the supportive words of Bill Langdon and Una-May O’Reilly encouraging me to project my voice and to push the boundaries of research in GP. It is impressive to see that collegiality still exists in our community twenty years on.
 |
| EuroGP98 Paris (pictured l-r) Conor Ryan, Mike O'Neill, Bill Langdon, Riccardo Poli, Peter Nordin, ?, Marco Tomassini, Una-May O'Reilly, Stefano Cagnoni, Frank Francone |
 |
| GP98 Madison, Wisconsin (pictured l-r) Mike O'Neill & J.J. Collins |
Since those early days I’ve presented many papers, posters, and tutorials. With great pride I've watched my own students present, and eventually became involved behind the scenes. This took the shape of the organisation of GECCO Workshops on Grammatical Evolution, Chairing the Graduate Student Workshop, and Chairing the GP and Real World Applications tracks to nowadays sitting on the SIGEVO Executive Board.
However, the highlight of my GECCO journey to date was having the honour of hosting GECCO 2011 in my home town of Dublin. This was a mammoth undertaking, which would not have been possible without the amazing team of researchers which make up our Natural Computing Research & Applications Group at University College Dublin.
In 2011 we welcomed nearly 700 submissions, and just shy of 600 delegates from 52 countries to Ireland. To assist us in making it all possible we also won much appreciated financial support (€36K) for GECCO from Science Foundation Ireland and the Irish tourist board Fáilte Ireland.
Recognising the significant work involved in hosting a GECCO, it was with a new found appreciation that I had the pleasure to immerse myself as a delegate in the impeccably organised GECCO in Berlin.
However, the highlight of my GECCO journey to date was having the honour of hosting GECCO 2011 in my home town of Dublin. This was a mammoth undertaking, which would not have been possible without the amazing team of researchers which make up our Natural Computing Research & Applications Group at University College Dublin.
In 2011 we welcomed nearly 700 submissions, and just shy of 600 delegates from 52 countries to Ireland. To assist us in making it all possible we also won much appreciated financial support (€36K) for GECCO from Science Foundation Ireland and the Irish tourist board Fáilte Ireland.
Recognising the significant work involved in hosting a GECCO, it was with a new found appreciation that I had the pleasure to immerse myself as a delegate in the impeccably organised GECCO in Berlin.
 |
Local Organising Team GECCO 2011 held in Dublin, Ireland
From left to right: Wei Cui, Kerry Sheehan, Eoin Murphy, Michael O’Neill, Jonathan Byrne, Cliodhna Tuite, Irene Ward, Erik Hemberg, Mike Fenton, Miguel Nicolau, Alex Agapitos, John Mark Swafford, Tony Brabazon and Dave Fagan
|
GECCO 2017 involved nine parallel sessions of tutorials and workshop over the first two days. Needless to say one can only be in one place at a time so my personal programme over the first two days involved taking in the EvoSoft Workshop (well just over the last hour as my plane only landed at 10am!), Deb’s tutorial on Recent Advances in Evolutionary Multi-Criteria Optimisation (EMO), and the Genetic Improvement Workshop.
EvoSoft saw the launch of PonyGE2 our groups python library for Grammatical Evolution (I arrived just after the presentation, sorry Mike!), which has evolved from what was originally a minimal implementation of GE (PonyGE) led by Erik Hemberg and James McDermott. The simplicity and ease of use of the Python language, and our groups immersion in a Business Analytics teaching environment (e.g., Python has amazing libraries for data manipulation, optimisation and Machine Learning) has resulted in the majority of our team adopting PonyGE2 in recent years. It was great to see Una-May O’Reilly and Sean Luke lead the debate during Q&A time, and then for Una-May to turn the tables on Sean Luke with some tough questioning when he got up to present his latest plans for one of the longest running libraries in our community, ECJ. Sean recently won some enviable funding from the NSF to develop out ECJ to embrace a wider set of metaheuristics, and we look forward to this enhanced flexibility.
Deb’s tutorial was a tour de force of recent advances in Evolutionary Multi-Criteria Optimisation, covering the development from multi-objective optimisation to the so-called many objective optimisation algorithms. It was excellent to hear how research priorities in this space are increasingly moving towards the decision making end of the process. Ok so I have a set of solutions comprising the pareto front, now which one do we use?
Deb noted how the many objective algorithms do not degenerate well to perform on 2 or 3 objectives, instead recommending to turn to the likes of NSGA-II in those instances. However, he hinted at some new work coming from their group which might provide an algorithm to unify multi and many objective optimisation, so keep a close watch on his group at MSU. Speaking of MSU and BEACON, over the past year I’ve had occasion to bump into Erik Goodman and Betty Cheng in Ireland, and it was a pleasure to catch up both with Erik and Wolfgang Banzhaf (and his son Benjamin).
Wolfgang has recently taken up the John R. Koza Chair for Genetic Programming at the BEACON Center in MSU. BEACON is growing from strength to strength with the high profile appointments of both Deb and Banzhaf and their cool approach of looking at both biological evolution and evolution in-silico (EC to you and I).
Finally, in the tutorial Deb highlighted the next instance of the EMO Conference which will celebrate its 10th Anniversary in Michigan during March 2019, so get your submissions ready for the deadline in Sep 2018 (you can’t say you haven’t been warned).
Deb noted how the many objective algorithms do not degenerate well to perform on 2 or 3 objectives, instead recommending to turn to the likes of NSGA-II in those instances. However, he hinted at some new work coming from their group which might provide an algorithm to unify multi and many objective optimisation, so keep a close watch on his group at MSU. Speaking of MSU and BEACON, over the past year I’ve had occasion to bump into Erik Goodman and Betty Cheng in Ireland, and it was a pleasure to catch up both with Erik and Wolfgang Banzhaf (and his son Benjamin).
Wolfgang has recently taken up the John R. Koza Chair for Genetic Programming at the BEACON Center in MSU. BEACON is growing from strength to strength with the high profile appointments of both Deb and Banzhaf and their cool approach of looking at both biological evolution and evolution in-silico (EC to you and I).
Finally, in the tutorial Deb highlighted the next instance of the EMO Conference which will celebrate its 10th Anniversary in Michigan during March 2019, so get your submissions ready for the deadline in Sep 2018 (you can’t say you haven’t been warned).
The remainder of my weekend was spent in the Genetic Improvement Workshop, where one of our researchers, Brendan Cody-Kenny presented a position paper on some blue sky thinking for GI and GP. Our group have developed an interest in EC for Search-based Software Engineering in recent years with some new funding from Science Foundation Ireland, and industry collaborator, Fidelity Investments, who have an Innovation Hub and undertake financial services software development in Ireland. This funding flows through Lero, the Irish Software Research Centre and supports research we are undertaking in genetic improvement. I was delighted to see Wolfgang’s invited talk where he raised, as it turned out, the not so controversial point of, the close relationship between Genetic Improvement and Genetic Programming, and also raised the possibility and promise of adopting developmental approaches in this domain. There were some standout talks at the workshop including those by Shin Yoo on Embedding GI into Programming Languages, Justyna Petke on new operators, David White on GI in No Time, and Saemundur Haraldsson on Fixing Bugs in Your Sleep.
In addition to the three excellent GECCO Keynotes, there were eight parallel sessions over the remaining three days, and my priorities were to take in as much of the Genetic Programming track as possible and the HUMIES, whilst dipping into a random selection of sessions and ensuring to support the various presentations and posters by researchers from our own group. The keynote by Drew Purves and Chrisantha Fernando from DeepMind generated a lot of interest, presenting a fascinating perspective on moving towards Artificial General Intelligence (AGI) through a combination of evolution, ecology and deep learning, and along the way they highlighted some of their recent progress with PathNet.
Hod Lipson’s keynote on Curious and Creative Machines reflected on some inspiring research in Genetic Programming for the automation of scientific discovery and self-reflecting robots, and he suggested optimistically that we won’t be replaced entirely through this automation as human scientists will be required to "collect the data, identify the building blocks and interpret the models"!
Additional highlights for me from a Genetic Programming perspective came in the form of Krzysztof Krawiec’s paper on formal specifications and Counterexample-driven GP (Winner of the GP Track Best Paper Award), Helmuth, McPhee & Spector's paper on improving generalisation through automatic simplification, and Moraglio & Krawiec’s paper on recursive boolean problems, which starts to push geometric semantic GP beyond regression towards application to more general programming tasks.
Hod Lipson’s keynote on Curious and Creative Machines reflected on some inspiring research in Genetic Programming for the automation of scientific discovery and self-reflecting robots, and he suggested optimistically that we won’t be replaced entirely through this automation as human scientists will be required to "collect the data, identify the building blocks and interpret the models"!
Additional highlights for me from a Genetic Programming perspective came in the form of Krzysztof Krawiec’s paper on formal specifications and Counterexample-driven GP (Winner of the GP Track Best Paper Award), Helmuth, McPhee & Spector's paper on improving generalisation through automatic simplification, and Moraglio & Krawiec’s paper on recursive boolean problems, which starts to push geometric semantic GP beyond regression towards application to more general programming tasks.
Behind the scenes, at a board meeting (over lunch) of the Genetic Programming & Evolvable Machines journal we learned about the healthy level of submissions, the relatively quick turnaround time (typically under 60 days) from submission to first decision, and the journals increasing impact factor.
Of particular welcome are the plans to increase the number of Peer Commentaries (see issue 18(3) for the latest in the series on Mapping of Genotype to Phenotype in EA's), which promote debate amongst researchers in the community on sometimes controversial topics. If you have an idea for a seed article for peer commentary that would prompt a healthy debate please do reach out to a member of the Editorial Board including yours truly.
Following GECCO 2011 in Dublin I was invited to put my name forward for election to the SIGEVO Executive Board, and was honoured to have been subsequently elected by members of the SIG. Much discussion amongst the board takes place over emails during the year, so GECCO presents an opportunity for the board to meet face to face, and this year was my first opportunity to attend.
It was exciting to hear the plans for GECCO 2018 in Kyoto. 15-19 July 2018 coincide with the Gion Festival, and the local organisers were keen to impress upon us the importance of booking a hotel room now, yes no kidding, now! as the room rates will rise substantially as the festival gets closer. There was also some discussion on the longer term plans for GECCO’s return to Europe and the American continent in subsequent years.
The new SIGEVO Summer School which bookend’ed GECCO was reported to be a resounding success with demand significantly outstripping the places available. Needless to say plans are being made to run the Summer school again in 2018.
A report on SIGEvolution, the newsletter of our SIG, was made and it is fantastic to see it return to it’s normal frequency of publication in recent times. With the current Editor Emma Hart taking up the position of Editor-in-Chief of the journal Evolutionary Computation a transition to a new editor for SIGEvolution will take place over the coming months, and congratulations are due to Gabriela Ochoa on stepping up to this important role. Gabriela was the Editor in Chief of GECCO 2017 and in the closing session of the conference she presented some wonderful analysis of the event since 2005 including a GECCO authorship network which is worth taking a look at.
Of particular welcome are the plans to increase the number of Peer Commentaries (see issue 18(3) for the latest in the series on Mapping of Genotype to Phenotype in EA's), which promote debate amongst researchers in the community on sometimes controversial topics. If you have an idea for a seed article for peer commentary that would prompt a healthy debate please do reach out to a member of the Editorial Board including yours truly.
Following GECCO 2011 in Dublin I was invited to put my name forward for election to the SIGEVO Executive Board, and was honoured to have been subsequently elected by members of the SIG. Much discussion amongst the board takes place over emails during the year, so GECCO presents an opportunity for the board to meet face to face, and this year was my first opportunity to attend.
It was exciting to hear the plans for GECCO 2018 in Kyoto. 15-19 July 2018 coincide with the Gion Festival, and the local organisers were keen to impress upon us the importance of booking a hotel room now, yes no kidding, now! as the room rates will rise substantially as the festival gets closer. There was also some discussion on the longer term plans for GECCO’s return to Europe and the American continent in subsequent years.
The new SIGEVO Summer School which bookend’ed GECCO was reported to be a resounding success with demand significantly outstripping the places available. Needless to say plans are being made to run the Summer school again in 2018.
A report on SIGEvolution, the newsletter of our SIG, was made and it is fantastic to see it return to it’s normal frequency of publication in recent times. With the current Editor Emma Hart taking up the position of Editor-in-Chief of the journal Evolutionary Computation a transition to a new editor for SIGEvolution will take place over the coming months, and congratulations are due to Gabriela Ochoa on stepping up to this important role. Gabriela was the Editor in Chief of GECCO 2017 and in the closing session of the conference she presented some wonderful analysis of the event since 2005 including a GECCO authorship network which is worth taking a look at.
 |
| HUMIES 2017 Bronze Award winners led by Mike Fenton (right) and joined by Mike O’Neill (centre) and Erik Hemberg (left) with the other co-authors (Ciaran McNally, Jonathan Byrne and James McDermott) back in Ireland. |
We were humbled to be in the company of the other shortlisted candidates including Robin Harper (winner of the Gold Award) who is finishing up his second PhD, which is in Quantum Physics. Robin's first PhD was in EC and involved Grammatical Evolution! His Gold winning entry saw a combination of his two areas of specialisation to impressively explain quantum correlations, or as Robin so nicely phrased it to discover the "boundary of reality”.
As a cycling enthusiast one of the downsides of GECCO for me is that each year it clashes with Le Tour, and I have to confess to keeping tabs on the cycling which is taking place in France over the duration of the conference. My excuse for missing GECCO 2015 in Madrid was that my family gave me a lovely present (for a significant birthday) to cycle a stage of the Tour (L’Etape du Tour). So in cycling speak “Chapeau”* to Robin Harper and the other HUMIES winners Shin Yoo et al (Silver) and Miikkulainen et al (Bronze), and a special “Chapeau” to all involved in the organisation of GECCO 2017.
 |
| Mike during GECCO 2015 at 2,067metres altitude on Col de la Croix de Fer, and proudly sporting his club (Bray Wheelers) colors! |
Comments
Post a Comment